Semigroup object additions

[igstan]

Adresses: #4615

[satorg]

It seems this PR was overlooked... Where do we stand on this?
Does it make sense to add some tests for the new methods added?

As I can see, there are some tests for the former static methods in SemigroupSuite. Perhaps, those tests don't look exactly comprehensive, but it's better than nothing.

[johnynek]

I guess adding it costs little, although the style this represents (which we used in Algebird at twitter to simplify type inference) isn't used outside of algebra (which adopted it from algebird)

That said, since those files are using it, I can see the argument for being consistent and I think we are unlikely to add more methods to the traits needed many more updates.

I would be for it to improve consistency with existing code.

[johnynek]

I don't see a huge benefit here since to use it you will often need Semigroup.reverse[String] compared to Semigroup[String].reverse which isn't a huge difference. You could avoid the type by passing to a method that expects a certain type, which may be common (since you are passing to an implicit/using in many cases).

[igstan]

Thanks. The reason I had added it was for consistency with the other methods.

In any case, I've removed the commit adding reverse and kept just the one adding intercalate.

[johnynek]

We could say the rule is: if the method takes a parameter of that is sufficient to infer the type of the result we add this method. In that way, it's still consistent.

[igstan]

True. I think that's a pretty good rule. Thanks for your review.